Large Language Models (LLMs) have achieved significant success in various tasks, yet concerns about their safety and security have emerged. In particular, they pose risks in generating harmful content and vulnerability to jailbreaking attacks. To analyze and monitor machine learning models, model-based analysis has demonstrated notable potential in stateful deep neural networks, yet suffers from scalability issues when extending to LLMs due to their vast feature spaces. In this paper, we propose ReGA, a model-based analysis framework with representation-guided abstraction, to safeguard LLMs against harmful prompts and generations. By leveraging safety-critical representations, which are low-dimensional directions emerging in hidden states that indicate safety-related concepts, ReGA effectively addresses the scalability issue when constructing the abstract model for safety modeling. Our comprehensive evaluation shows that ReGA performs sufficiently well in distinguishing between safe and harmful inputs, achieving an AUROC of 0.975 at the prompt level and 0.985 at the conversation level. Additionally, ReGA exhibits robustness to real-world attacks and generalization across different safety perspectives, outperforming existing safeguard paradigms in terms of interpretability and scalability. Overall, ReGA serves as an efficient and scalable solution to enhance LLM safety by integrating representation engineering with model-based abstraction, paving the way for new paradigms to utilize software insights for AI safety. Our code is available at https://github.com/weizeming/ReGA.

Reinforcement learning algorithms typically necessitate extensive exploration of the state space to find optimal policies. However, in safety-critical applications, the risks associated with such exploration can lead to catastrophic consequences. Existing safe exploration methods attempt to mitigate this by imposing constraints, which often result in overly conservative behaviours and inefficient learning. Heavy penalties for early constraint violations can trap agents in local optima, deterring exploration of risky yet high-reward regions of the state space. To address this, we introduce a method that explicitly learns state-conditioned safety representations. By augmenting the state features with these safety representations, our approach naturally encourages safer exploration without being excessively cautious, resulting in more efficient and safer policy learning in safety-critical scenarios. Empirical evaluations across diverse environments show that our method significantly improves task performance while reducing constraint violations during training, underscoring its effectiveness in balancing exploration with safety.

Safe motion planning algorithms are necessary for deploying autonomous robots in unstructured environments. Motion plans must be safe to ensure that the robot does not harm humans or damage any nearby objects. Generating these motion plans in real-time is also important to ensure that the robot can adapt to sudden changes in its environment. Many trajectory optimization methods introduce heuristics that balance safety and real-time performance, potentially increasing the risk of the robot colliding with its environment. This paper addresses this challenge by proposing Conformalized Reachable Sets for Obstacle Avoidance With Spheres (CROWS). CROWS is a novel real-time, receding-horizon trajectory planner that generates probalistically-safe motion plans. Offline, CROWS learns a novel neural network-based representation of a spherebased reachable set that overapproximates the swept volume of the robot's motion. CROWS then uses conformal prediction to compute a confidence bound that provides a probabilistic safety guarantee on the learned reachable set. At runtime, CROWS performs trajectory optimization to select a trajectory that is probabilstically-guaranteed to be collision-free. We demonstrate that CROWS outperforms a variety of state-of-the-art methods in solving challenging motion planning tasks in cluttered environments while remaining collision-free. Code, data, and video demonstrations can be found at https://roahmlab.github.io/crows/